📊 DPC Clustering Experiment Results

Reproduction of "Clustering by fast search and find of density peaks"
📄 Paper: "Clustering by fast search and find of density peaks"
👥 Authors: Alex Rodriguez and Alessandro Laio
🏛️ Journal: Science 344, 1492 (2014)
📈 Algorithm: Density Peaks Clustering (DPC)
🎯 Key Idea: Cluster centers are characterized by high density and large distance from points with higher density
Figure 2: Synthetic Data with 5 Density Peaks
Synthetic data with 5 density peaks of varying shapes and densities. Tests the algorithm's ability to detect non-spherical clusters with different densities.
Clustering Results
fig2_clusters.png
Shows the final clustering results. Different colors represent different clusters. Gray 'X' marks are halo/noise points. Yellow stars are cluster centers.
Decision Graph
fig2_decision_graph.png
Decision graph plotting δ (minimum distance to higher density point) vs ρ (local density). Cluster centers appear in the top-right corner (high δ and high ρ).
Gamma Values
fig2_gamma.png
Gamma values (γ = ρ × δ) sorted in descending order. Clear clusters show a sharp drop in γ values after the true centers.
Figure 3A: Two Crescent Moons
Two crescent-shaped clusters (moons dataset). Tests ability to detect non-convex clusters.
Clustering Results
fig3a_clusters.png
Shows the final clustering results. Different colors represent different clusters. Gray 'X' marks are halo/noise points. Yellow stars are cluster centers.
Decision Graph
fig3a_decision_graph.png
Decision graph plotting δ (minimum distance to higher density point) vs ρ (local density). Cluster centers appear in the top-right corner (high δ and high ρ).
Gamma Values
fig3a_gamma.png
Gamma values (γ = ρ × δ) sorted in descending order. Clear clusters show a sharp drop in γ values after the true centers.
Figure 3B: 15 Overlapping Clusters
15 highly overlapping Gaussian clusters. Tests resolution in distinguishing many clusters.
Clustering Results
fig3b_clusters.png
Shows the final clustering results. Different colors represent different clusters. Gray 'X' marks are halo/noise points. Yellow stars are cluster centers.
Decision Graph
fig3b_decision_graph.png
Decision graph plotting δ (minimum distance to higher density point) vs ρ (local density). Cluster centers appear in the top-right corner (high δ and high ρ).
Gamma Values
fig3b_gamma.png
Gamma values (γ = ρ × δ) sorted in descending order. Clear clusters show a sharp drop in γ values after the true centers.
Figure 3C: Three Concentric Circles
Three concentric circular clusters. Tests ability to detect nested structures.
Clustering Results
fig3c_clusters.png
Shows the final clustering results. Different colors represent different clusters. Gray 'X' marks are halo/noise points. Yellow stars are cluster centers.
Decision Graph
fig3c_decision_graph.png
Decision graph plotting δ (minimum distance to higher density point) vs ρ (local density). Cluster centers appear in the top-right corner (high δ and high ρ).
Gamma Values
fig3c_gamma.png
Gamma values (γ = ρ × δ) sorted in descending order. Clear clusters show a sharp drop in γ values after the true centers.
Figure 3D: Three Curved Clusters
Three curved, non-linearly separable clusters. Tests performance on complex shapes.
Clustering Results
fig3d_clusters.png
Shows the final clustering results. Different colors represent different clusters. Gray 'X' marks are halo/noise points. Yellow stars are cluster centers.
Decision Graph
fig3d_decision_graph.png
Decision graph plotting δ (minimum distance to higher density point) vs ρ (local density). Cluster centers appear in the top-right corner (high δ and high ρ).
Gamma Values
fig3d_gamma.png
Gamma values (γ = ρ × δ) sorted in descending order. Clear clusters show a sharp drop in γ values after the true centers.
Comparison: All Test Cases
Side-by-side comparison of DPC algorithm performance on all test cases from Figure 3.
Algorithm Performance Comparison
Comparison
Summary plot showing DPC clustering results on all four test cases from Figure 3.
Report generated: 2025-12-07 06:26:20